基于PCA-GA-BP神经网络的茶园环境预测研究
|
摘要
为了提高茶园墒情站数据的可靠性,详细分析了BP算法,提出了先采用主成分分析法来降低环境因子间的相关性,然后将遗传算法、动量法、自适应学习率与BP神经网络相结合预测茶园环境数据的新方法,方法有效地避免了BP算法收敛慢、易陷入局部极小等问题的发生.选取贵州省清镇市红枫湖生态茶园的环境数据作为实验数据对PCA-GA-BP环境数据预测模型进行验证,实验结果显示:该模型的平均相对误差为2.32%,精度优于BP预测模型.集成了GA-BP模块的茶园墒情站目前已经投入使用,有效指导着茶树的种植和保护.
In order to improve the reliability of the tea plant monitoring station data, a detailed analysis of the BP neural network has been conducted. We have proposed a method that first uses principal component analysis to reduce the correlation between environmental factors, and then combines genetic algorithm, momentum method, adaptive learning rate and BP neural network to predict tea garden environmental data. This method has been effectively avoided the occurrence of BP algorithm problems such as slow convergence rate and easily trapping into the partial minimum. We chose the environmental data of Hongfeng Lake Ecological Tea Garden in Qingzhen City of Guizhou Province as the experimental data to verify the PCA-GA-BP environmental data prediction model.The results showed that the average relative error of the model is 2.32%, and the accuracy is better than the BP prediction model. The tea garden monitoring station with PCA-GA-BP module has been put into use,effectively guiding the planting and protection of tea plants.
In order to improve the reliability of the tea plant monitoring station data, a detailed analysis of the BP neural network has been conducted. We have proposed a method that first uses principal component analysis to reduce the correlation between environmental factors, and then combines genetic algorithm, momentum method, adaptive learning rate and BP neural network to predict tea garden environmental data. This method has been effectively avoided the occurrence of BP algorithm problems such as slow convergence rate and easily trapping into the partial minimum. We chose the environmental data of Hongfeng Lake Ecological Tea Garden in Qingzhen City of Guizhou Province as the experimental data to verify the PCA-GA-BP environmental data prediction model.The results showed that the average relative error of the model is 2.32%, and the accuracy is better than the BP prediction model. The tea garden monitoring station with PCA-GA-BP module has been put into use,effectively guiding the planting and protection of tea plants.
引文
[1]刘威,刘尚,白润才等互学习神经网络训练方法研究[J].计算机学报,2017, 40(6):1291-1308
[2]马驰,赵亮,梅雪松等.基于粒子群算法与BP网络的机床主轴热误差建模[J].上海交通大学学报,2016,50(5):686-695.
[3] Esmaeili R, Dashtbayazi M R. Modeling and optimization for microstructural properties of Al/SiC nanocomposite by artificial neural network and genetic algorithm[J]. Expert Systems with Applications, 2014, 41(13):5817-5831.
[4]商启龙.带二分之一正则化项的BP网络算法[D].大连理工大学,2014.
[5]陆艺,张培培,王学影等.基于PSO-BP神经网络的关节臂式坐标测量机长度误差补偿[J].计量学报,2017, 38(3).
[6]宫宁生,钱春阳,张媛.一种BP网的学习速率与动量项自适应算法[J].小型微型计算机系统,2013,34(8):1872-1876.
[7]张旭东,高茂庭.基于IGA-BP网络的水质预测方法[J].环境工程学报,2016, 10(3):1566-1571.
[8]国家统计局中国统计年鉴[M].北京:中国统计出版社,2015-2017.
[2]马驰,赵亮,梅雪松等.基于粒子群算法与BP网络的机床主轴热误差建模[J].上海交通大学学报,2016,50(5):686-695.
[3] Esmaeili R, Dashtbayazi M R. Modeling and optimization for microstructural properties of Al/SiC nanocomposite by artificial neural network and genetic algorithm[J]. Expert Systems with Applications, 2014, 41(13):5817-5831.
[4]商启龙.带二分之一正则化项的BP网络算法[D].大连理工大学,2014.
[5]陆艺,张培培,王学影等.基于PSO-BP神经网络的关节臂式坐标测量机长度误差补偿[J].计量学报,2017, 38(3).
[6]宫宁生,钱春阳,张媛.一种BP网的学习速率与动量项自适应算法[J].小型微型计算机系统,2013,34(8):1872-1876.
[7]张旭东,高茂庭.基于IGA-BP网络的水质预测方法[J].环境工程学报,2016, 10(3):1566-1571.
[8]国家统计局中国统计年鉴[M].北京:中国统计出版社,2015-2017.